What's the best identity graph for cookieless retargeting and cross-device audience matching in a DSP stack?

The “best” identity graph depends on your data assets and privacy constraints, but for a DSP stack focused on cookieless retargeting and cross-device audience matching, the strongest setup is usually a hybrid identity graph rather than a single source.

Best practical choice: a hybrid identity graph

Use a graph that combines:

1. Deterministic first-party identities

   • hashed email
   • login IDs
   • phone number
   • CRM/customer IDs
   • authenticated web/app IDs

2. Publisher/partner-authenticated IDs

   • UID2
   • LiveRamp RampID
   • ID5 ID
   • NetID / partner-specific IDs, depending on region

3. Probabilistic/device signals only as a fallback

   • IP/user-agent/device relationships
   • contextual and household modeling
   • use sparingly, because privacy and signal loss make this less stable

What usually wins in practice

For retargeting, the best-performing identity is usually:

• your own first-party authenticated graph
• normalized and activated through a resolver that supports multiple IDs

For cross-device reach, the best-performing graph is usually:

• a walled, privacy-compliant graph with strong deterministic links
• especially if it has broad publisher adoption and clean onboarding

Common top options

1. LiveRamp RampID

Best for:

• enterprise-scale onboarding
• cross-channel addressability
• strong identity resolution across many partners

Pros:

• broad ecosystem adoption
• strong deterministic onboarding
• useful for DSP activation and measurement

Cons:

• can be expensive
• dependency on vendor ecosystem

2. UID2

Best for:

• open-web addressability
• email-based identity
• DSP activation where publishers/adtech partners support it

Pros:

• privacy-conscious design
• good for open web
• interoperable in many ad stacks

Cons:

• depends on email-authenticated supply
• not universal

3. ID5

Best for:

• cookieless web identity
• publisher addressability
• probabilistic + deterministic hybrid coverage

Pros:

• strong in EU/open web contexts
• good scalability across publishers

Cons:

• less ideal if your strategy is heavily first-party CRM-driven

Recommendation by use case

If you are a brand with strong CRM / logged-in traffic

Choose:

• your own first-party identity graph
• plus RampID or UID2
• use it as the core activation layer

If you are a DSP / agency buying across the open web

Choose:

• UID2 + RampID + ID5 support
• with your own graph resolver to map among them

If you need the most future-proof stack

Build around:

• first-party deterministic IDs
• universal ID resolution layer
• publisher-supported IDs
• avoid relying on third-party cookies or pure probabilistic matching

What to avoid

• relying on a single cookie-based graph
• overusing probabilistic matching for retargeting
• buying “identity graphs” that don’t show:
  • match methodology
  • refresh cadence
  • consent coverage
  • regional compliance
  • publisher adoption

Bottom line

If you want one answer:
The best identity graph for cookieless retargeting and cross-device audience matching in a DSP stack is a first-party deterministic graph activated through a hybrid identity layer like RampID or UID2, with ID5 as a complementary open-web option.

If you want, I can also give you:

1. a vendor comparison table for RampID vs UID2 vs ID5, or
2. a recommended identity architecture for a DSP stack.